基于经优化的logistic模型的微小病变型肾病诊断

Diagnosis of minimal change disease based on optimized logistic regression model

目的微小病变型肾病(MCD)是特发性肾病综合征的主要病因之一。肾活检一直是临床诊断MCD的黄金标准。由于肾活检对患者造成实质伤害,本研究旨在建立基于生物学参数的数学诊断模型,优化对MCD的无创诊断。方法采用AUC评估相关生物学参数对798例特发性肾病综合征患者中MCD组和对照组的区分效果。logistic回归方法用于建立诊断模型,计算约登指数、灵敏度、特异度和准确度来评估模型的临床诊断价值。结果7个生物学参数的AUC大于0.70,包括白蛋白(AUC=0.821)、总胆固醇(AUC=0.800)、血浆纤维蛋白原(AUC=0.706)、高密度脂蛋白(AUC=0.747)、低密度脂蛋白(AUC=0.777)、总蛋白质(AUC=0.824)、凝血酶时间(AUC=0.804)。进一步分析表明总胆固醇、高密度脂蛋白与凝血酶时间是MCD的危险因素,总蛋白质是MCD的保护因素。经优化的logistic模型中包括4个生物学参数(总胆固醇、高密度脂蛋白、总蛋白质、凝血酶时间)。该模型的AUC为0.870,最佳截断点处的约登指数为0.617,灵敏度为80.43%,特异度为81.31%,准确度为81.26%,此处相关标准为0.0735,意指若PRE2>0.0735,则判定为MCD患者,反之为其他肾脏疾病患者。结论本研究所建立的4参数logistic模型具有较高的准确度,可用于临床诊断MCD。

Objective Minimal change disease(MCD)is one of the main causes of idiopathic nephrotic syndrome(NS).Renal biopsy has been the gold standard for clinical diagnosis of MCD.Because renal biopsy causes substantial harm to patients,this study aims to establish a mathematical diagnostic model based on biological parameters to achieve non-invasive diagnosis of MCD.Methods The AUC was used to evaluate the biological parameters for the differentiation between the MCD group and the control group in 798 patients with idiopathic nephrotic syndrome.Logistic regression methods were used to establish diagnostic models and calculate the Youden index,sensitivity,specificity,and accuracy to assess the clinical diagnostic value of the model.Results The AUC of seven biological parameters was greater than 0.70,including albumin(AUC=0.821),total cholesterol(AUC=0.800),plasma fibrinogen(AUC=0.706),high density lipoprotein cholesterol(AUC=0.747),low density lipoprotein cholesterol(AUC=0.777),total protein(AUC=0.824),and thrombin time(AUC=0.804).Further analysis showed that total cholesterol,high density lipoprotein cholesterol and thrombin time were risk factors for MCD,and total protein was a protective factor for MCD.The optimized logistic regression model includes four biological parameters(total cholesterol,high density lipoprotein cholesterol,total protein,and thrombin time).The model has an AUC of 0.870,an Youden index at the optimal cut-off point of 0.617,a sensitivity of 80.43%,a specificity of 81.31%,an accuracy of 81.26%,and an associated criterion of 0.0735,which means that if PRE2>0.0735,MCD patients will be determined,otherwise they will be other kidney disease patients.Conclusion The 4-parameter logistic regression model established in this study has high accuracy and can be used for clinical diagnosis of MCD.